What are XBRL Tagging Outliers?
XBRL tagging outliers are a machine-learning-powered feature based on historical, filed XBRL data that helps XBRL filers identify where tag selections might vary from their industry group. Similar to other data quality checks provided within the XBRL Generation experience, XBRL tagging outliers are used to help focus review of those differences where they can be reviewed for appropriateness based on the related disclosures. This feature currently runs on 10-K or 10-Q reports only.
The results are provided based on:
- Standard line item concepts are compared in relation to the disclosure topics they are associated with in your document (e.g. elements appearing in your Business Combination/Acquisition Disclosure)
- This combination is then compared to your industry group
- Outlier results are provided when those combinations are uncommon compared to the industry group
If the generation includes an outlier, you will see a new message in the XBRL Validation panel after generating, titled “Insight - Tagging Outlier.” This message will include the outlier concept and its related XBRL Outline location.
By selecting “Review Usages,” the XBRL Locate Concepts panel will identify related document locations. Additionally, the XBRL Outline will directly open to the concept flagged as an outlier.
After reviewing current tag selections and alternative choices, you can update to a new element or mark the selection as correct.
Frequently asked questions
Who will have access to the messages?
Any user with the XBRL Manager role will be able to see outliers.
What data is used to train the model?
Outliers are based on historically filed XBRL data. Workiva has analyzed and gathered data from financial reports tagged for U.S. GAAP reporting as part of this process and applied the extensive expertise of Workiva. Models may periodically be trained on new public data as data is available and appropriate.
Are you collecting and using our data for models? If so, what data specifically?
Workiva's models do not collect or utilize customer data.
Does our model use customer data, e.g. financial numbers like revenue or disclosure data?
These models do not currently use any customer-specific data other than that which is available from public filings. We will collect metadata about the contexts used (e.g. XBRL tags and outline information) in order to monitor and improve the suggestions these models make to our customers. We will not use customer financial data in these models.
Does our model use Personally Identifiable Information (PII)?
No. None of our models use PII.
Who Developed the Outlier Detection Capability?
This feature is developed entirely within Workiva. No external partners are involved.